Re: Nulls, integrity, the closed world assumption and events

dawn wrote:
> JOG wrote:
> > dawn wrote:
> > > I think we need to come up with some
> > > better understanding of and appreciation for a "wiki" or "web" data
> > > model, one that pre-dates the RM. Because it has been around for many
> > > years, there should be some good emperical data regarding the use of
> > > this model compared to the RM, but I have not seen it.
> >
> > This statement made me feel physically nauseous for the poor people I
> > know who have worked tirelessly and fruitlessly on graph data models.
>
> I know there have been some recently, particularly related to the
> Semantic Web (and reading further, I see you reference that), who might
> not be too happy with the fruits of their labors. I have not been a
> fan of the Semantic Web, but figured it was trying to solve problems
> that I was not. I'm talking about the use of a similar model (di-graph
> of trees) for "structured data" not something that has been deemed
> "semi-structured."
>
> > There is now an immense understanding of the 'web data model' and I
> > know numerous people in both business and academia who feel they have
> > completely and utterly wasted a decade of their life generating it in
> > the name of the Semantic Web (at which /millions/ have been thrown by
> > Hewlett Packard alone).
>
> Yes, a shame, and other than a few books, not where I would put my
> dollars either. I'm sure there have been gains on the document "side
> of the house" as well as set-backs, but, again, I'm focussed on rather
> standard data processing (although everyone thinks their industry or
> apps to be non-standard in some ways).
>
> > Perhaps it is not solely your fault that you are unaware that the Web
> > data model has proven to be a complete and utter failure
>
> I'm aware that the research and work with "semi-structured" data with
> OWL, RDF, etc. has not exactly born much fruit. So, I am not talking
> about the actual web, but the abstracted data model, a di-graph of
> trees. I don't know if you are referring to such attempted standards
> as RDF. I haven't been reading about this lately, so I don't even know
> if RDF is dead, but I figured it was when RSS diverged from it. I
> suspect you are referring to other research, but, in any case, people
> have been using the di-graph with trees on nodes model (by whatever
> name) for real data processing for decades while the semantic web is
> only the last decade or so and does not quite address the same "problem
> space."
>
> >- the politics
> > of funding requires not publishing results that destroy all of your job
> > security or future income streams, so people's /abject/ disillusionment
> > is substantially unreflected in publications.
>
> I very much understand this, as I do know a bit about this--enough to
> resist any attempts to fund my own research at all, desiring to be
> completely free from any external pressures (and, therefore, just
> engaging in such efforts on the side, instead of picking up knitting at
> my age).
>
> > Nonetheless the
> > theoretical discussion of its insufficiency should have been a good
> > place to start.
>
> Perhaps there are some papers related to the Semantic Web that are also
> appropriate to structured data that would help me understand just what
> has been proven about its insufficiency. In my blog, I showed how the
> RM was not necessary in
> http://www.tincat-group.com/mewsings/2006/01/naked-model.html nor
> sufficient in
> http://www.tincat-group.com/mewsings/2006/02/dont-suffer-impedance.html
> for software development. If there is anything written with a similar
> audience in mind (me, for example) that might explain why the model
> that I and others have been using (di-graph with trees on nodes) is
> significantly impaired for developing data-based software, I would be
> happy to read it.
>
> Some of the pieces I have picked up here are the concern about query
> language bias, constraints specified to the DBMS as they are in
> SQL-DBMS's and the fact that PICK (like MUMPS) "gives the developer
> enough rope to hang themselves", use of non-declarative languages for
> constraint-handling, ambiguity of NULL when used with 2VL, ... None of
> these is a show-stopper, but there is some weight to the aggregate, and
> I do wish to have an understanding of both the pros and cons of various
> models.
>
> Thanks. --dawn

I am not clear what a di-graph with trees on nodes is (I envision some sort of nested graph structure). But let me say my grievances with graph based models come from bitter experience of working on a commercial implementation of such a thing (node identifiers and all) and developing an associated query language, for about two years. Given irreducible tuples, confusing what we ended up terming a structural layer (di-graphs) with the logical layer (only possible using n-ary edges) became an intractable problem - and one that I now see everywhere graph models emerge. We tried damn hard to work round it's deficits, and eventually had to concede defeat. The experience taught me that an n-ary logical model with value only addressing is essential in my experience. I wish I had known what I have learnt since back then. Received on Mon Jan 22 2007 - 04:11:32 CET